crewai

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • Prompt Injection (MEDIUM): The skill configures agents to use tools like SerperDevTool and WebsiteSearchTool to fetch data from the internet based on user-provided topics. This untrusted data is then incorporated into the agent's context. * Ingestion points: agents.yaml and tasks.yaml. * Boundary markers: None used to separate instructions from fetched content. * Capability inventory: CrewAI orchestration of multi-agent tasks. * Sanitization: No validation or filtering of external data is present.
  • External Downloads (LOW): The skill requires the 'crewai' Python package. This is a standard library for agent orchestration and is treated as a low-risk dependency finding.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 12:49 PM